The «One over Many»
Argument for Propositions

El argumento del «uno sobre los muchos»
para las proposiciones

Esteban Withrington

The Graduate Center – The City University of New York

Recibido: 17/12/21 Aceptado: 29/07/22

RESUMEN

Los significados de enunciados y pensamientos se consideran comúnmente en la semántica filosófica como objetos abstractos, llamados «proposiciones», que explican cómo diferentes enunciados y pensamientos pueden ser sinónimos y que constituyen los principales portadores de la verdad. Argumento que los significados son propiedades naturales que juegan roles causales en el mundo, que el tipo de pensamiento del «uno sobre los muchos» que subyace a la caracterización de significados compartidos como objetos abstractos está equivocado y que el hecho que los enunciados y pensamientos tengan valores de verdad en virtud de sus significados no implica que los significados sean portadores de la verdad.

PALABRAS CLAVE

Significados, Proposiciones, Uno-sobre-los-Muchos, Objetos Abstractos, Naturalismo

ABSTRACT

The meanings of utterances and thoughts are commonly regarded in philosophical semantics as abstract objects, called «propositions», which account for how different utterances and thoughts can be synonymous and which constitute the primary truth-bearers. I argue that meanings are instead natural properties that play causal roles in the world, that the kind of «One over Many»

thinking underlying the characterization of shared meanings as abstract objects is misguided and that utterances and thoughts having truth-values in virtue of their meanings does not entail that meanings themselves are truth-bearers.

KEYWORDS

Meanings, Propositions, One-over-Many,
Abstract Objects, Naturalism

1. Introduction

   A commonly held view in the field of philosophical semantics regards the meanings or contents of utterances and thoughts as abstract objects, called «propositions», which are mind- and language-independent and exist outside the spatio-temporal world. Such entities allegedly are the primary truth-bearers and constitute the shared meanings of synonymous utterances and thoughts. While it is clear that various sentences and beliefs do have common semantic features that are responsible for their truth or falsity, it is far from clear that these common features involve abstract objects. Consider the various utterances and written inscriptions of ‘Snow is white’ in English, of ‘La nieve es blanca’ in Spanish, of ‘Der Schnee ist weiss’ in German, and the various instances of the belief that snow is white in the minds of different people. Surely all these concrete sentence and belief tokens have «something» in common: they all mean or have the content that snow is white. Moreover, these sentence and belief tokens surely are true in virtue of their meanings or contents –together with snow being white. But why should we regard meanings as abstract objects rather than worldly features of sentence and belief tokens? The view that meanings are abstract objects is usually motivated by the metaphysical conviction that, when various sentence and belief tokens have the same representational content, what is common to them is a relation to a single object that exists independently of them: the same thing that is said or believed on many occasions and which has the property of being true or false. I maintain that this conviction is mistaken.

   In section II, I elaborate on the standard characterization of the meanings of utterances and thoughts as abstract objects. In section III, I discuss the kind of «One over Many» thinking underlying such characterization and explore reasonable alternatives that can characterize meanings as worldly phenomena. In section IV, I argue that such alternatives are preferable because relations to abstract objects fail to explain how utterances and thoughts can have meanings. In section V, I additionally argue that there are compelling reasons to regard meanings as worldly natural phenomena rather than as abstract objects –based on the causal role that meanings play in the spatio-temporal world. Finally, in section VI, I argue that utterances and thoughts having truth-values in virtue of their meanings does not entail that meanings themselves are the primary truth-bearers, so there is no case against the claim that meaningful utterances and thoughts are truth-bearers in their own right.

   Perhaps the most popular argument in favor of propositions in the recent literature relies on an account of the logical form of content ascriptions. The main claim is that ascriptions like ‘S believes that p’ or ‘S fears that p’ have the same logical form as ‘A loves B’: the verb functions as a dyadic predicate and ‘that p’ works as a «referential singular term» standing for an object (Schiffer, 1992). If we grant that some such ascriptions are true, it follows that the so-called «propositional attitudes» such as believing are relations between persons and real objects, propositions, which are typically argued to be abstract. This argument is not the target of this paper, but I will say a few words about it before proceeding. First, it should be noticed that there are alternative non-propositionalist accounts of the logical form of content ascriptions. One promising alternative is that they have the same form as ‘A loves some B’, where ‘B’ functions as a predicate rather than a singular term, so that the ascription asserts a relation between a person and some object –e.g., a token belief– that has the semantic property specified by ‘that p’ (Devitt, 1996, pp. 56-57).1 Additionally, if the «singular term» account were correct, we should be able to substitute ‘the proposition that p’ for ‘that p’ in content ascriptions without changing what they mean. But ‘Sam fears that winter is coming’ cannot be properly paraphrased as ‘Sam fears the proposition that winter is coming’ (Moltmann, 2003, p. 82). Finally, even if the «singular term» account happened to be right about the logical form of content ascriptions, it would not follow that ascriptions refer to propositions qua mind- and language-independent abstract objects: they may merely purport to refer to abstract objects that do not exist, but serve a purpose as useful fictions (Balaguer, 1998), or they may refer to naturalized propositions that derive their representational properties from those of mind and/or language (Soames, 2014; King, 2014; Hanks, 2015).

2. What are propositions alleged to be?

   Propositions are often said to be «what is common to a set of synonymous declarative sentences», so that two sentences «express the same proposition if they have the same meaning» (Haack, 1978, pp. 76-77). For example, ‘Snow is white’ in English, ‘La nieve es blanca’ in Spanish and ‘Der Schnee ist weiss’ in German have the same meaning, so they are all alleged to express the same proposition –namely, that snow is white. A proposition is supposed to be something separate from each of the declarative sentences that expresses it. It is what Alonzo Church calls a «proposition in the abstract sense» (Church, 1956a, p. 3). The word ‘proposition’ was traditionally used to refer to a «declarative sentence taken together with its meaning» (ibid., p. 3); but in its contemporary use the word refers only to the «content of meaning» of a declarative sentence (ibid., p. 5) which, according to Church, is an «abstract object» (Church, 1956b, p. 26). According to this view, what is common to synonymous sentences is that they are all related to a single abstract object.

   The claim that propositions are what synonymous declarative sentences have in common faces complications in the case of context-sensitive sentences. Presumably when Mary and John each say ‘I am hungry’, they are expressing different propositions, even though their utterances have, in a sense, the same meaning. So, as David Kaplan (1989) argues, a distinction needs to be drawn between two kinds of meaning. One kind is the unvarying meaning of an expression –e.g., utterances of ‘I am hungry’ always mean that the speaker is hungry– which Kaplan calls its «character». The other kind is the varying meaning of an expression that changes from context to context –e.g., ‘I am hungry’ means that Mary is hungry in one context and that John is hungry in another– which Kaplan calls its «content». Kaplan argues that, while the character of an expression depends on linguistic conventions, the content of a particular utterance of the expression depends on its character together with the context of the utterance. For example, the content of Mary’s utterance of ‘I am hungry’ is that Mary is hungry, because the character or linguistic convention for ‘I’ is that it refers to the speaker and the speaker in this context of utterance is Mary. Proponents of propositions can handle context-sensitivity by specifying that two utterances express the same proposition when they have the same content, rather than the same character. Kaplan himself identifies propositions with the contents of sentence utterances: «The content of a sentence in a given context is... a proposition.» (Kaplan, 1989, p. 500).

   Propositions are said to also be the contents of beliefs, identified in reports of the form ‘S believes that p’, where the ‘that’-clause contains a sentence expressing the content of the belief. Bertrand Russell, for example, said: «The content of a belief, when expressed in words, is the same thing... as what in logic is called a «proposition»… «That all men are mortal», «that Columbus discovered America»... are propositions.» (Russell, 1921, pp. 240-241). Russell characterized believing as a particular kind of psychological attitude that may be taken towards different propositions. Someone may believe that p, that q, that r, etc. Conversely, he pointed out that «there are various different attitudes that may be taken towards the same content.» (ibid., p. 243). Someone may believe that p, expect that p, remember that p, hope that p, fear that p, etc. It is worth pointing out that Russell rejected the view that propositions are entities that exist separately from concrete representations by the time he characterized beliefs as propositional attitudes (ibid., p. 241). Contemporary accounts of propositional attitudes, however, often adopt the view that they are relations between agents and abstract objects. According to this view, what the various tokens of the belief that snow is white in the minds of different people have in common is that they are all related to a single abstract object: the proposition that snow is white –the same object that constitutes the shared meaning of tokens of the sentence ‘Snow is white’ and its synonymous counterparts.

   Finally, propositions are claimed to be the primary bearers of truth. The idea is that particular sentences and beliefs are only derivatively true: a belief is true only insofar as it has a true proposition as its content and a sentence is true only when it expresses a true proposition. Talk of sentences or beliefs being true is considered at best as parasitic. Gottlob Frege, for example, claimed: «when we call a sentence true we really mean its sense is» (Frege, 1956, p. 292). The view is popular among contemporary philosophers. Some argue even that it is implicit in our ordinary way of talking. Paul Horwich, for example, claims: «Ordinary language suggests that propositional truth is fundamental and that the notion of an utterance ‘expressing a true proposition’ and the notion of a belief ‘being directed at a true proposition’ are understood in terms of it.» (Horwich, 1998, p. 133).

   To sum up, propositions are typically alleged to be abstract objects that constitute the common meanings or contents of synonymous sentences and beliefs and that are the primary truth-bearers. The main source of this view is Frege, who argued that propositions are «neither things of the outer world nor ideas», but entities that exist in a «third realm» (Frege, 1956, p. 302). But the view is still popular. Stephen Schiffer, for example, characterizes propositions as «abstract, mind- and language-independent objects» (Schiffer, 1992, p. 506), while George Bealer points out approvingly that this remains the «dominant view» among contemporary philosophers (Bealer, 1998, p. 2).

3. Meanings and the «one over many» argument

   An argument that seems to be implicit in the standard characterization of the common contents of synonymous sentence and belief tokens as abstract objects relies on the classical «One over Many» problem. I shall present the problem, discuss how it applies to meanings and argue that there are reasonable alternatives to the solution that postulates abstract objects.

   How can many things have one single feature in common? Multiple objects, for example, are red. This mundane situation may seem, as David Armstrong puts it, «profoundly puzzling»:

   The same property can belong to different things... Apparently, there can be something identical in things which are not identical. Things are one at the same time as they are many. How is this possible? (Armstrong, 1978, p. 11).

   An influential solution to this problem is transcendent realism: when many concrete and particular things share a common feature, F, it is because they are all related to one abstract and universal thing, F-ness, which exists independently of any of them. So, if many things are red, they must all be related to a single abstract object, redness itself, which would exist even if there were no red things. According to this view, «universals exist separated from particulars» (ibid., p. 140) and in a transcendent realm, independent from the spatio-temporal world of ordinary things (Armstrong, 1989, p. 76). The doctrine has been traditionally attributed to Plato.2 A contemporary proponent is Russell (1912, ch. 9).

   Notice that the same kind of reasoning can be applied to the semantic features of utterances and thoughts: we need to explain how numerically different sentence and belief tokens can have a single meaning or content in common that is responsible for their truth or falsity, and the solution proposed by transcendent realism is to postulate an abstract object that all of these tokens are related to. This transcendent form of «One over Many» thinking seems to underlie the standard characterization of the shared semantic properties of synonymous utterances and thoughts as abstract objects. Consider, for example, the following passage by George Pitcher:

   If one person says «It is raining,» another «Il pleut,» and a third «Es regnet,» a correct answer to the question «What did he say?» would in each case be «He said that it is raining» –for each would have said the same thing. And it is this element which all three utterances have in common –this same thing that is said in all three cases– that is the real bearer of truth, not the different sentences which the speakers happen to utter. (Pitcher, 1964, p. 5).3

   Since different synonymous sentence tokens all say the same thing, the passage suggests, this common feature must itself be another thing that they all are related to and that is the «real» truth-bearer. But this argument, as Richard Kirkham points out, is a «non sequitur» (Kirkham, 1992, p. 64). It simply does not follow from different utterances saying the same thing that each utterance does not say it by itself or that it is not a truth-bearer in its own right: «The [different] tokens say the same thing, but each one says it independently. Why could not each of the sentence tokens be true because of what it says?» (ibid., p. 64). The metaphysical principle that the common features of worldly things must involve a common relation to an abstract thing that «really» has that feature is itself problematic:

   Consider the property of being my sister: there are two women in the world who possess this property, and it is not something they have in common that «really» possesses the property, they each have it. There is not one thing that is my sister, there are two. (ibid., p. 65).

   Similarly, when various utterances say the same thing, there is no reason to suppose that what is said is another «thing», in addition to the utterances, that is the «real» truth-bearer. Kirkham’s reply works well against the kind of transcendent realism that seems to be implicit in Pitcher’s passage: one that assumes that F-ness itself must be F. A transcendent realist can reject this assumption to avoid the problems associated with it (Armstrong, 1978, pp. 71-72). The difficulties I discuss below, however, apply to any kind of transcendent realism.

   If we accept that a solution to the «One over Many» problem requires universals, there is a reasonable alternative to transcendent realism: we may adopt an immanent realism, which «admits universals but denies that they are transcendent» (ibid., 1978, p. 137). This view rejects that the common features of particular objects are themselves objects with a separate existence and proposes that, while universal features are real, they only exist in particular objects. The doctrine is commonly ascribed to Aristotle.4 Its main contemporary advocate is Armstrong (1978). According to immanent realism, redness is a universal but concrete property that exists in –rather than separately from– particular red things, while being a sister of Richard Kirkham is a worldly property possessed by two particular individuals, not an abstract entity that they both are related to. Similarly, it may be argued, the common semantic properties of various sentence and belief tokens are universal but worldly properties that exist in these tokens. Armstrong himself proposes an immanent realism about semantic properties, which avoids treating meanings as abstract objects (Armstrong, 2004, pp. 12-14).

   Another reasonable alternative is to be skeptical about the need to posit universals to begin with. Is there really any mystery regarding how different tokens have the same meaning or how different things can be red? Those who think that the «One over Many» problem is genuine find mundane situations like various things being red deeply puzzling and in need of explanation. Yet, Quine argues that this is a pseudo-problem and no explanation is needed:

   One may admit that there are red houses, roses and sunsets, but deny... that they have anything in common... That the houses and roses and sunsets are all of them red may be taken as ultimate and irreducible, and it may be held that [the realist about universals] is no better off, in point of real explanatory power, for all the occult entities which he posits under such names as ‘redness’. (Quine, 1961, p. 10).

   Quine’s suggestion is that particular things being red can be taken as a basic fact that needs no explanation and, additionally, that positing redness as a universal does not explain anything not already accounted for by acknowledging the existence of red things (ibid., p. 10). His point is that there is nothing puzzling to begin with and the alleged explanations are idle. Now, suppose that I say ‘My car and my bike have the same property: redness’. Defending the Quinean point of view, Michael Devitt argues that a claim like this can be paraphrased into a claim that serves the same purpose but is not committed to a universal: I can say instead ‘My car and my bike are both red’, which trivially follows from ‘My car is red’ and ‘My bike is red’ (Devitt, 2010, pp. 14-15). The realist about universals, however, is puzzled even about how a single thing can have a property: how can my car be red? But is this really puzzling? The fact that my car is red can be said to merely involve the existence of an object, my car, that really is red (ibid., p. 16). A further explanation in terms of universal redness seems to be neither required nor informative.

   Devitt acknowledges that there are legitimate explanations we may seek for something having a property, but he argues that they are not related to the problem of universals: we may wonder what caused or what is the purpose of something having certain property, and, in the case of a non-fundamental property, we may wonder what constitutes something having that property (ibid., p. 16). Strictly speaking, my car and my bike being red are facts that need explanation. But suppose that we have a satisfactory reductive explanation of what constitutes my car and my bike being red in terms of what wavelengths of light they each reflect, as well as a satisfactory causal explanation of why each of them does reflect certain wavelengths. It looks like in such a case we have managed to explain everything that needed explanation. The realist about universals will not be satisfied, since the problem of universals concerns even how objects can have the most fundamental properties. But perhaps she is looking for an additional explanation that would serve no genuine purpose: «The explanation must stop somewhere. What better place to stop than with a fundamental physical fact of our world?» (ibid., p. 17).

   These considerations can be applied to the question of how different sentence and belief tokens can have the same meanings. A token of ‘Il pleut’ and a token of ‘Es regnet’ having the same semantic property, it may be argued, is just a matter of both of them meaning that it rains: neither transcendent nor immanent universals are required. While a token meaning that it rains is a fact in need of reductive explanation, the legitimate search for such an explanation is not related to the problem of universals. Explanations must stop somewhere and no further explanation is required once we have a satisfactory naturalized semantics that explains what fundamental properties constitute meanings.

   I have argued that there are reasonable alternatives to the transcendent solution to the «One over Many» problem for synonymous sentence and belief tokens. One option is an immanent realism according to which meanings are universal but worldly properties that exist in the tokens. Another option is to reject the «One over Many» as a pseudo-problem. Neither of these alternatives posits abstract objects. Yet they are typically ignored when abstract objects are introduced as what is common to synonymous tokens: a transcendent realism seems to be simply assumed.

   In fairness, there are even more alternatives to transcendent realism: various forms of nominalism attempt to solve the «One over Many» problem while denying the existence of universals. Just to mention a couple of traditional proposals: predicate nominalism claims that «for particulars to have the same property… is for the same predicate to apply to them» (Armstrong, 1978, pp. 138-139), while resemblance nominalism explains things having a «common property» in terms of «the resemblance which the particulars bear to each other» (ibid., p. 44). These views face prima facie objections. It looks like the predicate ‘red’ applies to various things because they are red, rather than the other way around; otherwise, things could not be red if the predicate did not exist (ibid., p. 17). Similarly, it looks like red things resemble each other because they are red, rather than the other way around; otherwise, a thing could not be red if no other red things existed (ibid., pp. 50-53). These forms of nominalism seem to invert the correct order of explanation. The proposal by Quine and Devitt is also a form of nominalism, since it rejects realism about universals. But it does not face such problems: my car and my bike being both red needs neither an explanation in terms of universals nor one in terms of resemblance or predicate application. This view does not invert the correct order of explanation and allows for something to be red even if there are no other red things or there are no predicates applying to red things.

   A more recent alternative is particularism about properties or trope nominalism, which claims that «the properties of particulars are themselves particular» instead of universal (ibid., p. 59). According to this view, the redness of my car and the redness of my bike are two numerically different particular properties or «tropes». How can this view explain that various things are red if they have numerically different properties? The main suggestion appeals to resemblance between tropes (Williams, 1953; Campbell, 1990). Armstrong initially complained that this faces the same prima facie objections as resemblance nominalism (Armstrong, 1978, pp. 84-85). However, trope nominalism does not explain the particular redness of my car by a resemblance to the particular redness of any other thing, but the other way around: «the resemblances do flow from the natures of the resembling things», as Armstrong later conceded (Armstrong, 1989, p. 44). The view should not be so easily dismissed. Notice that, according to a trope nominalism about meaning, each token utterance and thought has its own particular semantic property and the synonymy between tokens is a matter of resemblance between their semantic tropes. This provides another alternative that avoids treating meanings as abstract objects that exist outside the spatio-temporal world.

4. Problems with relational analyses

   I have argued that there are reasonable alternatives to transcendent realism about meanings. But why should we prefer any of these alternatives over transcendent realism? An important reason is that transcendent realism fails to explain how things can have shared properties because, as Armstrong shows, it generates a vicious relation regress.

   Transcendent realism explains how various particular objects have the same property in terms of their relation to a transcendent universal. But it must also treat this relation as a transcendent universal, so explaining how the objects have this relation requires the same sort of explanation in terms of a further relation, which in turn needs to be explained by yet another relation, leading to an infinite regress (Armstrong, 1978, pp. 70-71). It follows that positing relations to abstract meanings fails to explain how various sentence and belief tokens can be synonymous and even how each of them has a meaning at all, since the attempted explanation generates a vicious regress. This is a well-known problem in the literature on the nature of properties, but it is seldom if ever addressed when meanings are characterized as abstract objects in philosophical semantics.

   Armstrong argues that any attempt to solve the problem of universals that appeals to a relational analysis is undermined by a relation regress, since «the relations appealed to by such analyses themselves require the same analysis» (ibid., p. 139). Traditional forms of nominalism analyze things having something in common in terms of relations such as resemblance or falling under a predicate. Particularism about properties also appeals to relations of resemblance between tropes to explain how different things have something in common. But such relations themselves need to be analyzed in terms of further relations, generating vicious regresses (ibid., pp. 18-21, 53-56 & 84-85).5 The problem even affects immanent realism: if an object having an immanent property is a relation it has to this property, then it must also have the immanent property of having this relation, generating a vicious regress (ibid., pp. 106-107).

   To overcome the regress problem, Armstrong proposes a non-relational immanent realism according to which objects and their properties have an «intimate union» that is not a relation: the properties of a thing are not «separate constituents of the thing» that «must then be related» to it (ibid., pp. 107-108). But if properties are not separate constituents of things, how can different things have the same universal property? Armstrong admits that this is «literally inexplicable»: «We simply have to accept that different particulars may have the same property...» (ibid., p. 109). The proposal also seems unable to explain how a single thing can have a property. Devitt objects that Armstrong merely «declares» the union to be non-relational, but we have no idea what this could involve, so that Armstrong «replaces the explanatory failings of relational Realism with a complete mystery» (Devitt, 2010, p. 18). If all other attempts fail and Armstrong’s own proposal is not a genuine solution, this provides support to the view that the «One over Many» is a pseudo-problem (ibid., pp. 18-19).

   In later writings, Armstrong accepts that the «fundamental tie» between things and their properties may be relational and generate an infinite regress, but suggests that it is not a vicious regress because all the further relations supervene on the fact constituted by the first relation (Armstrong, 1989, pp. 109-110). The further relations become an «ontological free lunch» (ibid., p. 100). This misses the point that the problem of an infinite regress is that the explanation of an object having a property cannot be accomplished. If further relations are ontologically harmless because they supervene on the first one, they cannot also explain the first one without circularity.

   Advocates of trope nominalism have also argued that their resemblance-based solution to the «One over Many» can avoid a vicious regress. One option is to rely merely on resembling tropes, denying that their resemblances are further tropes, while another option is to admit tropes of resemblance but deny that the regress they generate is vicious (Maurin, 2002, ch. 5). The latter option seems to face the same problem as Armstrong’s supervenience strategy. But trope nominalism faces a more fundamental regress problem in its explanation of how an object can have tropes –which in the mainstream version that treats objects as bundles of tropes involves the relation of «compresence» between tropes. Perhaps the best attempt to solve this problem is the one that treats the relation as a compresence trope, but denies that it generates a vicious regress because it is relata-specific and dependent upon its relata (Maurin, 2010). This account seems to be circular, explaining how tropes are bundled by a compresence trope and, at the same time, explaining the compresence trope by the tropes it bundles (Briceño, 2016, p. 67; Imaguire, 2018, p. 71).

   The jury is still out on whether any solution to the problem of universals can avoid a vicious regress. There is a growing literature on the issue that I cannot address properly here. Some version of immanent realism or trope nominalism may overcome the problem. But the difficulties facing all attempts may indicate that no solution can avoid it. After a thorough review of the literature, Guido Imaguire argues that all attempts are unsatisfactory and we need an approach along the lines of Quine’s and Devitt’s nominalism (Imaguire, 2018, ch. 4).

   The sort of nominalism proposed by Quine and Devitt is immune to any relational regress objections, since it holds that no explanation, relational or not, is required. Armstrong dismissively calls this view «Ostrich Nominalism», complaining that it constitutes a «refusal to take predicates with any ontological seriousness» (Armstrong, 1978, p. 16). He thinks that we need F-ness to account for the truth of predications of the form ‘a is F’. The Quinean sees no such need, given the standard semantic account according to which ‘a is F’ is true if and only if «there exists an x such that ‘a’ designates x and ‘F’ applies to x» –where the value of x is «not a universal but simply an object» (Devitt, 2010, p. 16). Such object is not a featureless «bare particular» either: it must really be F for the predication to be true (ibid.). Recall that the view is not a form of predicate nominalism: it does not claim that an object is F because the predicate ‘F’ applies to it, but the other way around. Armstrong objects that this gives predicates «power without responsibility», because it entails that they carry «no ontological commitment» (Armstrong, 1980, pp. 443 & 445). But he is wrong. As Devitt points out, the account entails that my assertions of ‘Lulu is a cat’ and ‘Nana is a cat’ ontologically commit me to the existence of cats, since cats must exist for them to be true (Devitt, 2010, p. 22). Predicates do carry ontological commitment according to this view, although it is not the commitment that Armstrong expects.6 I conclude that Armstrong’s criticism is misplaced and the sort of nominalism proposed by Quine and Devitt is a reasonable alternative –and perhaps the most reasonable one, since it avoids the regress problems that seem to undermine all the other ones.

   I have argued that transcendent realism faces a relational regress that undermines its account of how things can have shared properties. So it cannot explain how utterances and thoughts have meanings in common. If the «One over Many» problem is genuine, perhaps a version of immanent realism or trope nominalism can solve it. Alternatively, the problem may not be genuine. While I sympathize with this latter view, my argument does not depend on it. All of the remaining alternatives agree that meanings are natural worldly phenomena.

5. The causal roles of meanings

   There is a compelling metaphysical reason to suppose that semantic properties of sentence and belief tokens are natural worldly properties of each of these tokens. Abstract objects cannot causally act upon any concrete things in the world, since «we have fairly good scientific reasons to believe that Nature, the spatio-temporal system, is a causally self-enclosed system» (Armstrong, 1978, p. 129). The causal closure of the physical world undermines, for example, the hypotheses that there is a God that acts upon the world or that we have non-physical souls that interact with our bodies. According to such hypotheses, some physical events have non-physical causes. But that violates the causal closure principle.

   The same considerations apply to the postulation of propositions qua abstract objects. Abstract objects cannot causally interact with concrete –datable and placeable– utterances and thoughts, so they cannot determine their semantic properties. As Devitt points out, abstract objects «can play no causal role in mind and language», so «we have the best of reasons for thinking that they are not part of mental and linguistic reality.» (Devitt, 1996, p. 210). Armstrong rejects the postulation of propositions qua abstract objects for the same reason:

   There are metaphysicians who are prepared to postulate a realm of propositions over and above the space-time world. But, presumably, we could not stand in any causal or nomic relation to such a realm. And if we cannot stand in such relations to propositions it is unclear that such a postulation is of any explanatory value. (Armstrong, 2004, p. 12).

   That propositions qua abstract objects cannot act upon the world is rarely disputed nowadays. But the transcendent realist about propositions may be willing to accept that they do not causally interact with concrete utterances and thoughts.

   Accepting that propositions do not causally interact with utterances and thoughts, while claiming that they constitute their meanings, amounts to regarding meanings as causally inert epiphenomena. This is deeply troubling, since we postulate meanings or contents in order to causally explain people’s behavior (Devitt, 1996, pp. 57-60).7 Ascribing to Mary a belief that it is raining, a belief that using an umbrella can help her not to get wet, and a desire not to get wet, plays a crucial role in the explanation of why she picks up an umbrella. Similarly, ascribing to Mary’s utterance of ‘It is raining’ the meaning that it is raining plays a crucial role in ascribing to her the belief that it is raining, which in turn plays a crucial role in explaining her behavior: why she picks up an umbrella. Had the contents of her beliefs, her desire and her utterance been different, we would reasonably expect her behavior to vary accordingly.

   The causal closure of the physical world provides a compelling reason for regarding the semantic properties we ascribe to explain behavior as natural worldly properties, rather than relations to abstract objects. Moreover, our explanations are often successful, which provides evidence that the meanings ascribed do play a causal role in the physical world. Consequently, the view that meanings are abstract, mind and language independent objects, must be rejected.

6. Should we discard or naturalize propositions?

   Granting that the contents of utterances and thoughts are worldly properties, should we reject propositions or attempt to provide a non-transcendent account of propositions?

   In Quinean fashion, Devitt prefers to avoid talk of propositions altogether, and talks instead of meaningful sentence and belief tokens:

   By talking of meaningful tokens, we can avoid talk of propositions. This is an advantage not simply because propositions are creatures of darkness but also because talk of propositions... is explanatorily unhelpful. (Devitt, 1996, p. 13).

   Devitt admits that propositional talk may be legitimate to the extent that such talk can be «paraphrased away, when the ontological chips are down, into talk about the properties of concrete thoughts and utterances» (ibid., p.210). Although he warns that, even then, propositional talk is «unnecessary and misleading» (ibid., p. 210). Presumably, it is unnecessary because it is an avoidable manner of speaking and it is potentially misleading because there is the risk of taking literally what is only useful as a figure of speech.

   Armstrong, in contrast, does not want to reject propositions altogether. Instead, he proclaims: «as a naturalist, I want to look for a this-worldly account of propositions» (Armstrong, 2004, p. 12). His proposal is that propositions are indeed the contents of beliefs and the meanings of sentences, but that they are properties to be explained by a naturalized semantics, rather than abstract objects: «I would be hoping for a naturalistic theory of content and meaning, and so a naturalistic theory of... propositions.» (ibid., p. 14).

   The difference between the approaches of Armstrong and Devitt may seem merely verbal. After all, they both regard meanings as natural properties. Yet, there is a genuine difference. Armstrong, unlike Devitt, argues that propositions –not sentence and belief tokens– are the primary truth-bearers (ibid., p.12). Devitt’s view is best characterized as treating meaningful tokens as the primary truth-bearers –tokens that are true or false in virtue of their semantic properties. Armstrong even proposes that propositions are abstractions but, he insists, «not in any other-worldly sense» (ibid., p. 13). He gives the following reason:

   That... content or meaning is an abstraction becomes clear when we notice that contents and meanings are types rather than tokens. Beliefs in different minds may have the very same content, numerically different statements may have the very same meaning. (ibid., p. 13).

   Basically, Armstrong thinks there is a «One over Many» problem for synonymy, so he suggests an immanent realism, according to which propositions are worldly but universal semantic property types that exist in sentence and belief tokens.

   The disagreement between Armstrong and Devitt regarding propositions stems from their disagreement on realism about universals and on whether the «One over Many» is a genuine problem. But why is there a need for propositions if Armstrong is right? Why not just talk about meanings (qua immanent properties)? Armstrong offers the following reason: «I am inclined to think that all other suggested truthbearers besides propositions are called truthbearers on account of their relationship to certain propositions.» (ibid., p. 12). In other words, he agrees with the need for «something» that is not only the common meaning of synonymous tokens, but also the primary bearer of truth. Such a «thing» is an immanent universal property, rather than an abstract object, but plays the same roles the alleged abstract object was expected to play. Should naturalized meanings play the same roles propositions qua abstract objects where expected to play?

   It is far from clear that meanings should play the role of primary truth-bearers once we have rejected the account of meanings qua abstract objects and replaced it with an account of meanings as worldly properties –whether universal or not. The semantic properties of sentence and belief tokens surely are partly responsible for their truth. A Spanish speaker’s utterance of ‘La nieve es blanca’ and someone’s belief that snow is white are true because they both mean that snow is white –and because snow indeed is white. But it does not follow from this fact alone that the real bearers of truth are meanings themselves, rather than meaningful sentence and belief tokens –that is to say, concrete utterances and thoughts with their semantic (and syntactic) properties. Meanings or contents, abstracted from utterances and thoughts, may not even be legitimate –let alone primary– truth-bearers. An analogy may help to clarify this point:

   A boat has the property of positive buoyancy –of weighting less than the water it displaces– and, due to this, it has the property of floating in water. But what floats is the boat, not the property that allows it to float. While other objects also float because they are positively buoyant, it does not follow that the property they have in common floats. Positive buoyancy does not float, positively buoyant objects float.

   Similarly, it may be reasonably argued:

   A token of ‘La nieve es blanca’ in Spanish has the property of being true because it has the property of meaning that snow is white –and snow indeed is white. But what is true is the token, not the property that allows it to be true. While there are other sentence and belief tokens that are true also because they mean or have the content that snow is white, it does not follow that the semantic property they have in common is true. Meanings or contents are not true or false, meaningful sentence and belief tokens are true or false.

   My point here is not that meanings cannot have the property of being true or false simply because they are properties. After all, some properties may themselves have properties. For example, the property of being red may have the property of being a color.8 What the analogy with positive buoyancy shows, rather, is that an object O having property X in virtue of having property Y does not entail that X must be a property of Y itself in order to be a property of O. So, truth-values being properties of meanings and, consequently, sentence and belief tokens being only derivatively true or false, does not follow merely from meanings being properties in virtue of which sentence and belief tokens can be true or false.

   To establish that meanings are the primary truth-bearers, we would need to show that the relation between meaning and truth is more like the relation between being red and being a color than like the relation between positive buoyancy and floating. Why should we regard being true as a second-order property of first-order meaning properties, rather than as a first-order property of meaningful tokens? Armstrong does not provide any reasons to think so. So he begs the question against the reasonable alternative that the primary bearers of truth are meaningful sentence and belief tokens. So I conclude that there is no support for the claim that naturalized meanings also play the roles of primary truth-bearers and deserve to be regarded as «propositions».

7. Conclusion

   I have discussed the view that the meanings or contents of utterances and thoughts are the abstract objects called «propositions» which constitute the primary truth-bearers. I have argued that the argument for propositions qua abstract objects based on the «One over Many» argument fails, since the transcendent «solution» faces an insurmountable difficulty: it faces a vicious relation regress that undermines its explanation of how different utterances and thoughts can be synonymous. We must either look for a non-transcendent solution or disregard the «One over Many» as a pseudo-problem. I additionally argued that meanings must be natural worldly properties, since they play causal roles in the explanation of behavior. This excludes abstract objects as prospective meanings or contents.

   We must either reject propositions altogether or look for a naturalistic account of propositions. I have questioned, however, the need for naturalized meanings playing the same roles as transcendent propositions, since there is no reason to believe that meanings themselves are truth-bearers. Meaningful sentence and belief tokens seem to be truth-bearers in their own right.

   VIII. References

   Aristotle (1908). Metaphysics. Translated by W. D. Ross. The Works of Aristotle. Volume VIII. Oxford: Clarendon Press.

   Armstrong, D. (1978). Nominalism and Realism: Universals and Scientific Realism. Vol. 1. Cambridge: Cambridge University Press.

   Armstrong, D. (1989). Universals: An Opinionated Introduction. Boulder: Westview Press.

   Armstrong, D. (1980). «Against ‘Ostrich Nominalism’: A Reply to Michael Devitt». Pacific Philosophical Quarterly, Vol. 61, Issue 4, 440-449.

   Armstrong, D. (2004). Truth and Truthmakers. New York: Cambridge University Press.

   Balaguer, M. (1998). «Attitudes without Propositions». Philosophy and Phenomenological Research, Vol. 58, No. 4, 805-826. https://doi.org/10.2307/2653723

   Bealer, G. (1998). «Propositions». Mind, Vol. 107, Issue 425, 1-32. https://doi.org/10.1093/mind/107.425.1

   Briceño, S. (2016). «El regreso de Bradley y el problema de la unidad-compleja: ¿Tropos al rescate?». CRÍTICA, Revista Hispanoamericana de Filosofía, Vol. 48, No. 143, 47-75. https://doi.org/10.22201/iifs.18704905e.2016.218

   Campbell, K. (1990). Abstract Particulars. Oxford: Basil Blackwell.

   Church, A. (1956a). «Propositions and Sentences». The Problem of Universals. Bocheński, Church and Goodman. South Bend: University of Notre Dame Press, 3-11.

   Church, A. (1956b). Introduction to Mathematical Logic. Princeton: Princeton University Press.

   Davidson, D. (1984). Inquiries into Truth and Interpretation. Oxford: Clarendon Press.

   Devitt, M. (1996). Coming to Our Senses: A Naturalistic Program for Semantic Localism. New York: Cambridge University Press.

   Devitt, M. (2010). Putting Metaphysics First. New York: Oxford University Press.

   Fine, G. (1993). On Ideas: Aristotle’s Criticism of Plato’s Theory of Forms. New York: Oxford University Press.

   Fine, G. (2003). Plato on Knowledge and Forms: Selected Essays. New York: Oxford University Press.

   Frege, G. (1956). «The Thought». Translated by A. M. and Marcelle Quinton. Mind, Vol. 65, No. 259, 289-311. https://doi.org/10.1093/mind/65.1.289

   Haack, S. (1978). Philosophy of Logics. Cambridge: Cambridge University Press.

   Hanks, P. (2015). Propositional Content. New York: Oxford University Press.

   Horwich, P. (1998). Truth. Second Edition. New York: Oxford University Press.

   Imaguire, G. (2018). Priority Nominalism: Grounding Ostrich Nominalism as a Solution to the Problem of Universals. Cham: Springer International Publishing.

   Kaplan, D. (1989). «Demonstratives». Themes for Kaplan. Almog, Perry and Wettstein (eds.). New York: Oxford University Press, 481-563.

   King, J. (2014). «Naturalized Propositions». New Thinking about Propositions. King, Soames and Speaks (eds.). New York: Oxford University Press, 47-70.

   Kirkham, R. (1992). Theories of Truth: A Critical Introduction. Cambridge, MA: The MIT Press.

   Lycan, W. (1988). Judgement and Justification. Cambridge: Cambridge University Press.

   Maurin, A.S. (2002). If Tropes. Dordrecht: Kluwer Academic Publishers.

   Maurin, A.S. (2010). «Trope Theory and the Bradley Regress». Synthese, Vol. 175, No. 3, 311-326. https://doi.org/10.1007/s11229-009-9511-2

   Moltmann, F. (2003). «Propositional Attitudes without Propositions». Synthese, Vol. 135, No. 1, 77-118. https://doi.org/10.1023/A:1022945009188

   Moltmann, F. (2017). «Cognitive Products and the Semantics of Attitude Verbs and Deontic Modals». Act-Based Conceptions of Propositional Content: Contemporary and Historical Perspectives. Moltmann and Textor (eds.). New York: Oxford University Press, 254-289.

   Pitcher, G. (ed.) (1964). Truth. Prentice-Hall: Englewood Cliffs.

   Plato (1991). The Republic. Translated by Allan Bloom. The Republic of Plato. Second Edition. New York: Basic Books.

   Quine, W.V.O. (1961). From a Logical Point of View. Cambridge, MA: Harvard University Press.

   Rodriguez-Pereyra, G. (2002). Resemblance Nominalism. A Solution to the Problem of Universals. Oxford: Clarendon Press.

   Russell, B. (1912). The Problems of Philosophy. Oxford: Oxford University Press.

   Russell, B. (1921). The Analysis of Mind. London: George Allen and Unwin.

   Sellars, W. (1963). Science, Perception and Reality. Atascadero: Ridgeview Publishing Company.

   Schiffer, S. (1992). «Belief Ascription». The Journal of Philosophy, Vol. 89, No. 10, 499-521. https://doi.org/10.2307/2941005

   Soames, S. (2014). «Cognitive Propositions». New Thinking about Propositions. King, Soames and Speaks (eds.). New York: Oxford University Press, 91-124.

   Williams, D.C. (1953). «On the Elements of Being I-II». The Review of Metaphysics, Vol. 7, Nos. 2-3, pp. 3-18, 171-192.

   Esteban Withrington es Doctor en Filosofía por The Graduate Center – The City University of New York.

   Líneas de investigación:

   Filosofía del Lenguaje, Filosofía de la Mente y Semántica Naturalizada.

   Publicaciones recientes:

   2019: Basic-Acceptance Teleosemantics. ProQuest Dissertations Publishing.

   Email: ewithrington@gradcenter.cuny.edu

© Contrastes. Revista Internacional de Filosofía, vol. XXVIII Nº1 (2023), pp. 61-79. ISSN: 1136-4076

Departamento de Filosofía, Universidad de Málaga, Facultad de Filosofía y Letras

Campus de Teatinos, E-29071 Málaga (España)